In the intrusive polynomial chaos expansion (PCE) based stochastic finite element method, a Galerkin projection translates the stochastic system into a coupled set of linear equations with deterministic coefficients. The size of the deterministic system is proportional to the finite element mesh resolution and the number of stochastic parameters. For high resolution models, efficient solvers are required to tackle this system. The solvers must be parallel and scalable to large number of processors in order to efficiently exploit the available computing power of supercomputers. Recently, various domain decomposition algorithms are reported for stochastic partial differential equations (SPDEs) using the polynomial chaos expansion (PCE). In this paper, the performance of the probabilistic versions of the dual-primal finite element tearing and interconnect method (FETI-DP) is investigated for three-dimensional problems using MPI and PETSc parallel libraries and METIS graph partitioning software.
展开▼
